Do we still need to introduce ChatGPT? At work, at school or at home, the tool developed by OpenAI has found its supporters and its critics. But if the revolutionary chatbot has scored points, the next version promises to be a great leap forward.
This Wednesday, the head of OpenAI, Sam Altman, returned to TechCrunch on the expectations surrounding GPT-4, the fourth version of the chatbot (ChatGPT being the public name of an improved version of GPT-3).
“It will come out when we are convinced that we can publish it safely and responsibly,” the manager immediately states, acknowledging that ChatGPT has wreaked havoc in the academic world with its significant risks of cheating.
One thing is certain, GPT-4 will not be an “artificial general intelligence” (AIG). This term refers to an artificial intelligence that approaches humans, especially for cognitive tasks. We are talking about creation, feelings and even consciousness… the Grail of AI researchers.
GPT-4 “will disappoint”
“I think an IAG is pretty much what’s expected of us,” but GPT-4 will “disappoint” people who have this expectation, insists Sam Altman. The latter, however, believes that IAG could appear faster than expected and without it being a sudden technological leap.
Still, GPT-4 should pass a new course. Last year, the head of a startup working with OpenAI predicted 100 trillion parameters for this new version against “only” 175 billion parameters for ChatGPT. Actually, it is mostly in the delicacy of the language and in its way of producing answers on more complex topics that will be expected.
Pending the arrival of GPT-4, OpenAI could also offer a new version of its Dall-E artificial intelligence imager and, above all, a first tool dedicated to video.
Source: BFM TV
